Liver transplantation analysis - Time to graft failure

This is the Code used for analysis of graft failure after liver transplantation. Unfortunately, the data could not be shared, therefore only the code is provided in this repository.

Preprocessing

• import_raw.R: imports raw data, performs basic data cleaning, variable renaming and non-model based pre-selection, creates new variables. Does not perform general plausibility checks.

• Note that transplantation date was only given with monthly precission.

• apply_exclusion.R: applies exclusion criteria, stores number of deleted observations per exclusion criterion.

• create_survvars.R: calculates the event/censoring time and status variable depending on which information is given.

• The above steps are summarized in run_preproc.R

creation of the survival time and event/censoring indicator

The calculation of survival times and event/censoring indicators is based on the following variables:

• retrans: When time until retransplantation is given we know the exact date of the procedure and know that the current liver did not completely fail until then, thus we consider this the most reliable information on the liver failure time

• death: When retrans and death are given, we use the retrans variable as time to failure, otherwise we use death.

• lastseen: When lastseen is given and nothing else, we consider lastseen as censoring time

• fail and reregis: for patients where both, fail/reregis and either death or retrans were given we have seen that these variables are not reliable as the time difference between them can be very large, thus we exclude observations where only fail or reregis are given

• When neither of the variables where given we set administrative censoring at 2013-12-31 and calculated the censoring time as 2013-12-31 - transplant.date.

Modeling

• The confounder model is fit in confounder_model.R using mgcv::gam
• The final model including donor age is fit in donorage_model.R